Introduction

All known organisms use ribosomes for protein synthesis. Consequently, the ability to independently synthesize proteins is one of the defining characteristics of life. Cells dedicate a considerable effort to the production and maintenance of ribosomes to allow protein biosynthesis. The majority of total RNA and protein synthesis in bacteria and the yeast Saccharomyces cerevisiae is attributed to the production of ribosomal RNAs (rRNA) and proteins1,2. During the course of evolution towards eukaryotes, cell size and complexity increased and the cells required an increasing number of ribosomes to meet the considerably larger demand for protein synthesis.

The eukaryotic ribosome consists of ~80 ribosomal proteins, which maintain the structure of the ribosome, and four different rRNAs, which are mainly responsible for protein synthesis. To produce more ribosomes, cells developed several unique systems. Production of ribosomal proteins can be achieved by boosting translation of their mRNAs, but synthesis of rRNA is more difficult and relies on optimization of ribosomal DNA (rDNA) transcription, which is performed by the highly specialized RNA polymerase I (Pol I)2. To further increase production of rRNA, cells developed dedicated systems for rRNA gene amplification and copy number maintenance3. Eukaryotic cells are thus able to stably retain more than 100 rRNA gene copies in their chromosomes and thus produce a large number of rRNA transcripts4. In the budding yeast, one of the best studied model organisms, ~2,000 ribosomes are produced per minute in the nucleolus2. The nucleolus has been described as the busiest ‘factory’ of the cell and its impairment affects most cellular functions5. Thus, we cannot understand the biology of the cell without studying ribosome biogenesis as a prerequisite for protein production.

In this Review, we discuss how cells stably produce the substantial amount of rRNA through rRNA gene amplification and specialized transcription. We focus on the recent developments in research of rDNA repeat structure, copy number maintenance, transcriptional and epigenetic regulation, and nucleolus organization. Finally, as rDNA is known to be an unstable region of the genome, we discuss the role of ribosome biogenesis in cell senescence and human diseases. With an increasing understanding of the nucleolus and the processes it is home to, a comprehensive picture of the dynamic production of rRNA as the initial step of ribosome biogenesis is beginning to emerge.

rDNA organization and maintenance

To facilitate mass production of rRNA, each cell contains many copies of the RNA gene in unique tandem clusters. However, due to their repetitive organization and active transcription, rDNA clusters are highly susceptible to loss of gene copies through inter-repeat recombination. To negate this process, cells have developed a sophisticated gene amplification system that allows maintenance of the correct rRNA gene number.

Organization and maintenance of rDNA clusters in yeast

Prokaryotes, such as Escherichia coli, have seven copies of rDNA, even though the genome is relatively small compared with that of eukaryotes1. The rRNA gene copies are scattered near the replication initiation site (oriC) in the same orientation in which replication proceeds in the circular chromosome. The origin-proximal position of the rRNA genes could be advantageous in allowing their duplication at early stages of the S phase. The orientation of transcription in the same direction of DNA replication prevents ‘head-to-head’ collision of rRNA synthesis with DNA replication6. As the level of rDNA transcription is very high compared with average gene transcription levels, the frequency of collisions between replication and transcription machineries could be very high and result in replication fork arrests, DNA double-strand breaks (DSBs) and genome rearrangements. Therefore, there seems to be selection pressure to retain the orientation of transcription with regard to replication7. Additionally, the DNA replication terminator protein Tus binds to 22-bp Ter sequences in the replication termination zone and blocks those replication forks that counteract transcription directionality8. Hence, rRNA gene copy number and position and orientation relative to the replication machinery are important to support rRNA production in bacteria.

Similar mechanisms are in place in eukaryotic cells, which require a greater number of ribosomes owing to their increased size compared with prokaryotes. Many eukaryotic cells carry more than 100 rRNA genes4. The budding yeast Saccharomyces cerevisiae contains roughly 150 copies of rRNA genes in chromosome XII (refs. 9,10) (Fig. 1a). Each repeat in the tandem array has a length of 9.1 kb and comprises two rRNA genes: the 5S and 35S rRNA genes11. The 5S and 35S rRNA genes are separated by two intergenic spacers (IGSs) — IGS1 and IGS2 — which contain regulatory elements for transcription, replication and rDNA maintenance. The 35S rRNA gene is transcribed by the dedicated Pol I and produces the 35S precursor rRNA, which is processed by removal of the internal transcribed spacers, the 5′ external transcribed spacer and the 3′ external transcribed spacer into three mature rRNAs: 18S, 5.8S and 23S RNAs (Fig. 1a). By contrast, the 5S rRNA gene is transcribed by Pol III (which transcribes also tRNAs and other short structured RNAs)12.

Fig. 1: rDNA organization and recombination in budding yeast.
figure 1

a, Organization of ribosomal DNA (rDNA) in the budding yeast Saccharomyces cerevisiae. About 150 copies of the ribosomal RNA (rRNA) gene are located as a tandem array in chromosome XII. The 35S rRNA genes (encoding the 23S, 5.8S and 18S rRNAs) are interspersed with 5S rRNA genes and with intergenic spacer 1 (IGS1) and IGS2. Arrows indicate the direction of transcription of rDNA and of the non-coding promoter E-pro, as well as the direction of replication from the ribosomal autonomously replicating sequence (ARS). b, During S phase, replication proceeds from the replication origin in a bidirectional manner; however, the fork moving against the direction of rDNA transcription is blocked at the replication fork barrier (RFB) by fork-blocking protein 1 (Fob1). DNA double-strand breaks (DSBs) can form at arrested replication forks. In cells with normal numbers of rDNA copies, E-pro transcription is repressed by the histone deacetylase Sir2 and cohesion rings restrict repair of DSBs to between equal sister chromatids. Reduction of rDNA copy number by deletion-causing recombination between rDNA repeats releases upstream activating factor (UAF) complexes from their binding sites at 35S promoters, and they are able to bind the SIR2 promoter and repress Sir2 expression. The resulting reduction in Sir2 activity allows activation of transcription from the E-pro locus, resulting in displacement of cohesin and recombination between the unequal sister chromatids, leading to gene copy amplification. Following recovery of normal rDNA copy numbers, UAF is again sequestered at rDNA promoters, SIR2 expression increases and E-pro is silenced28,34. In cells with low numbers of rDNA copies, repair through intrachromosomal homologous recombination may lead to production of extrachromosomal rDNA circles (ERCs). Cen, centromere; ITS, internal transcribed spacer; Tel, telomere. Figure adapted with permission from ref. 24, AAAS.

Full size image

Resulting from rDNA damage and its repair, deleterious homologous recombination between rDNA copies often occurs13. To avoid loss of repeats, cells have developed maintenance systems, which rely on sequence elements located in the IGS. A key element is the replication fork barrier (RFB), which is located at the end of the 35S rRNA gene14,15,16,17. The RFB is crucial for arresting the replication fork arriving from the replication initiation sequence known as the autonomously replicating sequence (ARS). As shown in Fig. 1a, replication forks proceed bidirectionally from the ARS, and the fork moving from left to right in the opposite direction of 35S RNA gene transcription is blocked at the RFB. The reason for this polarity is apparently to avoid collision between the DNA and RNA polymerases, as observed on the E. coli chromosome16,17. This process requires binding of fork-blocking protein 1 (Fob1) to the RFB18,19.

The RFB site also acts as a recombination hotspot by inducing DSBs20,21,22. Although repair of these DSBs can lead to loss of repeats, homologous recombination also offers the opportunity of increasing the copy number when the DSBs are repaired using as a template a neighbouring homologous region at the sister chromatid, resulting in amplification of rRNA gene copies10,23 (Fig. 1b). Another key factor in increasing rDNA copy number is ‘E-pro’, a bidirectional promoter of a non-coding RNA24,25,26. Active transcription from E-pro promotes dissociation of the cohesin complex24, which connects the sister chromatids following replication (Fig. 1b). Removal of cohesion enables homology-directed repair of DSBs through recombination with neighbouring rDNA copies (unequal sister chromatid recombination) and increases copy number. Interestingly, E-pro is exchangeable with the bidirectional promoter of GAL1/GAL10, but not with the unidirectional promoter of GAL7 (ref. 24), indicating that bidirectional transcription is required for E-pro functionality. The reliance of the process on transcription explains why the yeast strain in which E-pro was replaced by a galactose-induced GAL1/GAL10 promoter can increase its rDNA copy number only in galactose medium27. These mechanisms explain how increase of rDNA copy number can be accomplished in cells. However, amplification regulation is also required to maintain the proper copy number. When the copy number is reduced, E-pro transcription is activated to enhance recombination between unequal sister chromatids24. The resulting copy number increase restores the desired rDNA copy number and represses E-pro transcription.

The questions remains of how E-pro-dependent transcription itself is regulated and how cells sense the correct number of rDNA copies. The histone deacetylase Sir2 can act as a repressor of E-pro transcription in cells with regular rDNA copy numbers22 (Fig. 1b). In a strain with a low rDNA copy number, the amount of Sir2 is reduced, resulting in E-pro activation and thus copy number increase24. A genetic screen for factors that inhibit SIR2 expression identified upstream activating factor (UAF)28. UAF is a complex consisting of Uaf30, two copies of histone H3, histone H4 and the subunits Rrn5, Rrn9 and Rrn10, with the last three including one histone fold domain each29,30,31. UAF primarily associates with the promoter of the 35S rRNA gene to upregulate rDNA transcription29,32. Interestingly, UAF also binds to the SIR2 promoter and represses it, likely by preventing assembly of the Pol II preinitiation complex that is necessary for its transcription. As the amount of UAF is limited to ~20 molecules per yeast cell33, the factor can be released from rRNA genes only if their copy number is reduced, resulting in its migration to and repression of SIR2. The reduced SIR2 expression finally leads to E-pro activation, thereby inducing rDNA copy number amplification. Once wild type copy number levels (~150 copies) are restored, UAF leaves the SIR2 promoter to again bind rDNA promoter sequences and restore SIR2 expression. Consequently, the increase in Sir2 levels results in repression of E-pro-dependent transcription and rDNA amplification ceases28,34. This titration-based feedback system allows the efficient regulation of rDNA copy number (Fig. 1b).

In addition to rDNA amplification, recombination produces circular molecules, called ‘extrachromosomal rDNA circles’ (ERC), when the same sister chromatid is used as the template (Fig. 1b). ERCs accumulate in the yeast mother cells and reduce their lifespan (see later)35.

rDNA organization in mammalian cells

The organization of mammalian rDNA is similar to that of budding yeast, except that the 5S rRNA gene forms independent arrays in a different genomic locus (Fig. 2). The 47S rDNA loci are present on the short arms of acrocentric chromosomes (chromosomes 13, 14, 15, 21 and 22 in human cells) and form nucleolus organizer regions. Sequences flanking rDNA are called ‘distal junctions’ at the telomeric side and ‘proximal junctions’ at the centromeric side; the different distal junctions are conserved among chromosomes in humans36. A repeating rDNA unit is 45 kb long and includes the coding region producing the 47S precursor rRNA, which is processed into the three mature rRNAs: 18S, 5.8S and 28S rRNAs. Compared with their yeast counterparts, both the internal transcribed spacer (2,237 bp in humans and 593 bp in yeast) and the IGS (31 kb in humans and 2.5 kb in yeast) are much larger in human cells. The sequence similarity of coding regions between human cells and budding yeast is 75% in 18S rRNA, 73% in 5.8S rRNA and 51% in 25S/28S rRNA, whereas the internal transcribed spacer and IGS share almost no similarity (Y.H. and T. K., unpublished data).

Fig. 2: Organization of the ribosomal rRNA gene in human cells.
figure 2

a, Human 47S ribosomal RNA (rRNA) genes comprise the 18S, 5.8S and 28S ribosomal transcripts and are organized as clusters on acrocentric chromosomes 13, 14, 15, 21 and 22. As in yeast, the genes are separated by intergenic spacer (IGS) sequences. Transcription termination factor 1 (TTF1) arrests replication forks similarly to fork-blocking protein 1 (Fob1) in budding yeast, by associating with the Sal box sequence in the R repeat, which is also a terminator of 47S rRNA transcription. The function of butterfly/long repeats is not known. Arrows indicate the direction of rRNA gene transcription and of ribosomal DNA (rDNA) replication from the replication initiation zone. b, Human 5S rRNA genes. The repeat is located on chromosome 1. ITS, internal transcribed spacer.

Full size image

Although the mammalian IGS contains retrotransposons and repetitive sequences that largely account for its increased length, elements crucial for rDNA maintenance such as the RFB site, replication initiation zone and a non-coding RNA promoter are functionally conserved. The RFB site is located at the end of the 47S rRNA gene, and transcription termination factor 1 (TTF1; homologue of yeast Reb1 and Nsi1) associates with the 18-bp Sal box sequence in the R repeat. This R repeat is a terminator of the 47S rRNA transcription in the IGS. TTF1 binding to the R repeat terminates not only rDNA transcription but also the replication fork37,38 (Fig. 2a). TTF1 thus functions similarly to Fob1 in S. cerevisiae despite not showing any sequence homology. The exact position of the replication origin at mammalian rDNA has not yet been determined. As the RFB site functions to block the replication fork from entering the 47S rRNA gene in the opposite direction of transcription, it is likely that the origin is located downstream of the gene, namely in the IGS region. Until now, an E-pro-like promoter that enhances recombination has not been found in animals, even though some spacer promoters were identified in frogs and mice39,40,41. Although transcription from these spacer promoters seems to regulate 47S rRNA transcription, it may have a role also in recombination.

The mechanisms of amplification and copy number maintenance are not well understood in mammalian cells. As the copy number typically lies within a range of 200–700 repeats per cell in humans42,43, the existence of regulatory mechanisms that maintain this high copy number is likely. Similarly to in budding yeast, the rDNA repeat organization is rather regularly aligned in a manner such that the mostly identical rDNA sequences are repeating without irregularity42. The regular alignment of rDNA repeats suggests that recombination-mediated gene conversion occurs to homogenize the sequence as described in S. cerevisiae11, but mechanistic insight is lacking. In cancer cells and in DNA repair-deficient mammalian cells, rDNA instability can be detected in the form of copy number variation by pulsed-field gel electrophoresis44,45. In such recombinogenic cells, the RFB site may enhance instability through its propensity for DSB formation. Even in wild type cells, some recombination occurs at a certain rate and likely contributes to maintaining the optimal copy number and sequence homogeneity of rRNA genes.

rDNA transcription by RNA polymerase I

To accomplish the high transcription levels required for rRNA synthesis, the Pol I transcription system has accumulated unique features in the structures of its transcription factors, in its promoter DNA sequence and in the polymerase enzyme itself. Curiously, all genes in the nucleus are transcribed by Pol II or Pol III, with the sole exception of the rDNA genes, underlining the odd nature of the Pol I system. Recent structural and functional analyses of Pol I activity in vitro and its regulation in cells have improved our understanding of its adaptations to the unique task of synthesizing rRNA precursors in the nucleolus.

The general transcription factors of Pol I

The 13-subunit Pol I complex (14 subunits in Saccharomycotina species) uses a minimal set of general transcription factors throughout the transcription process. First, Rrn3 is recruited to Pol I and stabilizes an initiation-competent form of the enzyme46 (initially described as Pol Iβ) (Fig. 3). Rrn3 was identified in mouse (termed ‘TIF1A’)47, yeast48 and human49,50 cell lysate fractions containing initiation-competent Pol I. Targeting of the Pol I–Rrn3 complex to active rRNA gene promoters in yeast is achieved by the transcription factor complex UAF in cooperation with the core factor (CF) complex and TATA-binding protein (TBP)29. UAF is not conserved in mammals and is functionally replaced by upstream binding factor (UBF), despite both factors being highly divergent in their composition. Metazoan selectivity factor 1 (SL1) has characteristics of CF but includes additional subunits and TBP51,52. Transcription initiation by the yeast Pol I–Rrn3 complex requires only CF in vitro, but addition of UAF and TBP strongly stimulates basal transcription levels29. Promoters of rRNA genes do not share conserved sequences across species, but functional elements were identified in most model organisms53. A promoter DNA core element in S. cerevisiae stretches from the transcription start site 38 bp upstream; further up lies the upstream activating element, which specifically binds UAF54. Figure 3 outlines the key factors of the Pol I transcription process.

Fig. 3: Outline of the RNA polymerase I transcription process.
figure 3

RNA polymerase I (Pol I) can form a dimer in its inactive, ‘expanded’ conformation. Monomers are stabilized and prevented from inactivation by binding the transcription initiation factor Rrn3, which is a prerequisite for their recruitment to the ribosomal RNA (rRNA) gene promoter. In yeast, upstream activating factor (UAF) binds an element upstream of the promoter, thereby allowing the recruitment of TATA-binding protein (TBP) and the core factor complex. Transcription initiation requires a specific DNA conformation that allows binding of the Pol I–Rrn3 complex to core factor. Upon binding of the Pol I–Rrn3 complex to core factor, contraction of Pol I occurs (see also Supplementary Fig. 1b) and either causes or coincides with promoter DNA melting (not shown). Following promoter escape, Rrn3 is released and elongating Pol I fully contracts. High mobility group protein 1 (Hmo1) dimers bind across active rRNA genes, supporting transcript elongation. Transcription termination is caused by the proteins Reb1 (in fission yeast) or its homologue Nsi1 (in budding yeast), which are bound to a specific termination sequence and act as a ‘roadblock’ for Pol I. H3, histone H3; rDNA, ribosomal DNA.

Full size image

Pol I structure and function are optimized for pre-rRNA synthesis

The Pol I transcription system shows many similarities to Pol II, Pol III and archaeal or even bacterial systems, especially in RNA polymerase architecture and mechanism of catalysis55. However, it appears that Pol I accumulated structural and functional adaptations to its highly dedicated task. Compared with transcription of the rDNA gene by Pol II, transcription of the rDNA gene by Pol I occurs at very high loading rates56, is faster (higher polymerization rates) on long templates57 and features intrinsic proofreading capacity by cleavage of mismatched product RNA58.

Comparison of high-resolution Pol I crystal structures59,60 with their Pol II counterpart61 demonstrated that Pol I carries built-in transcription factors, which resemble the tuneable Pol II-interacting factors. Firstly, the Pol I subunit RPA12 represents a hybrid form of Pol II subunit RPB9 and cleavage factor TFIIS62; it increases elongation efficiency63 and allows intrinsic RNA-cleavage activity during proofreading and backtrack release64. The Pol I subunit complex RPA49–RPA34 resembles Pol II initiation factors TFIIF and TFIIE, but is tightly associated with the Pol I core65 (Supplementary Fig. 1a). A linker domain of RPA49 aids stabilization of the initial transcription bubble and positions the carboxy-terminal (C-terminal) tandem winged helix domain of subunit RPA49. This tandem winged helix domain is required to achieve high loading rates66 and assists transcription initiation67, likely by supporting promoter escape68,69.

Additional Pol I-specific domains (Supplementary Fig. 1) include the ‘connector’ and ‘expander’ elements. The latter can occupy the active centre cleft to prevent unspecific transcription by Pol I dimers59,60 or monomers70 by mimicking the DNA template, thereby providing a safeguard mechanism against unregulated activation. The connector of subunit RPA43 mediates the formation of inactive Pol I dimers71 in S. cerevisiae, explaining the lack of transcription activity in experiments in high molecular weight lysate fractions that apparently contained such dimers72. An early structure of inactive Pol I dimers was obtained by negative stain electron microscopy73 and was only later fully understood. Dimerization may be a way of storing inactive, ‘hibernating’ Pol I molecules during starvation periods71. Interestingly, a recent study demonstrated Pol I dimerization independent of the connector domain in another yeast70, but has not yet been reported in more complex eukaryotes thus far. Pol I is also adapted to allow detection of UV-induced DNA damage. The residue Arg1015 of its largest subunit in S. cerevisiae, RPA190, is highly specific for Pol I and is responsible for transcription stalling at bulky lesions in rDNA, resulting in enhanced activation of transcription-coupled repair compared with Pol II (ref. 74).

Interestingly, the incorporation of nucleotides is both faster and more promiscuous in Pol I compared with Pol II (refs. 75,76). The reason may lie in an evolutionarily optimized balance between speed and error rate and may reflect the unique cellular functions of each enzyme.

Transcription initiation by Pol I is optimized for high efficiency

Similarly to Pol I itself, some transcription factors share structural and functional similarities with their Pol II and Pol III counterparts, but specialized to their functions in pre-rRNA synthesis. This is especially true for the transcription initiation factors. Initiation machineries are comparable between Pol II and Pol III, but diverge in Pol I in terms of composition and structures of factors77.

In yeast, promoter targeting of Pol I and prevention of Pol II transcription are both achieved by UAF, which binds to a sequence element upstream of the promoter and enhances Pol I recruitment in cooperation with TBP29,32 (see earlier). A recently determined cryogenic electron microscopy (cryo-EM) reconstruction of S. cerevisiae UAF reveals a particle of six histone fold domains, to which the upstream promoter DNA is bound, inducing DNA bending (rather than wrapping as seen for nucleosomes), and shows TBP in a sequestered state31. TBP increases the efficiency of Pol I initiation in cooperation with UAF, but basal levels of transcription are achieved independently of both factors29. Hence, the specific activity and structural dynamics of UAF and TBP in targeting Pol I to its promoter and enhancing initiation remain to be clarified. Nevertheless, studies on the basal mechanisms of CF-mediated targeting of the Pol I–Rrn3 complex to the core promoter element revealed that biophysical DNA properties (meaning a combination of ‘bendability’ and ‘meltability’) and its positioning relative to the transcription start site may be more important than the promoter sequence itself78,79.

Upon transcription initiation, Pol I contraction allows the formation of the enzyme’s active centre77,80, as illustrated by cryo-EM reconstructions of actively elongating polymerases in vitro81,82 and by cryogenic electron tomography of actively transcribing enzymes on rRNA genes81. Two different theories based on these structural and functional studies have been postulated to explain the fast and efficient promoter melting and initiation of Pol I. DNA ratchetting upstream of Pol I has been proposed to drive promoter melting, based on the study of Rrn3-free initially transcribing complexes68. Another study suggested trapping of stretches of single-stranded template and non-template DNA upon Pol I contraction in an easy-to-melt promoter region following upstream-DNA bending and steric distortion of duplex DNA, which is a more likely explanation83. Both mechanisms are possible if one takes into account the location of initial bubble formation84 and the analysis of promoter regions in other model organisms85. To allow entry of upstream promoter DNA into the central cleft deeper in Pol I (ref. 35) compared with Pol III (refs. 86,87) or Pol II (refs. 88,89), and to prevent canonical TBP–DNA binding90, the CF component Rrn7 exhibits several adaptations in comparison with its counterparts TFIIB (in Pol II) or Brf1 (in Pol III systems). By contrast, TFIIB and Brf1 share many similarities in structure and mechanism of action91,92. Harnessing these adaptations, Pol I molecules couple CF-dependent recruitment of Pol I–Rrn3 complexes with immediate promoter melting, template and non-template DNA strand trapping, transcription initiation and promoter escape, thus drastically reducing promoter dwell time. In line with this hypothesis, closed promoter initiation complexes are under-represented in single-particle cryo-EM analyses due to their inherently transient nature. Such efficient initiation at high rates appears to be required to sustain high Pol I loading rates, which is crucial for optimal rRNA synthesis in exponentially growing cells93.

Although the mechanisms of Pol I function have been extensively studied over the past few years, especially in budding yeast, many questions remain. Defined positions of the Pol I-specific ‘expander’ element and Pol I subunit RPA12 were observed in transition states following Pol I recruitment69,79, but it remains unclear whether they are actively involved in transcription initiation or whether their presence in structural intermediates is coincidental. Deletion of the expander59,60 does not impair cell growth, and a direct functional requirement for the RPA12 C terminus in transcription initiation has not yet been demonstrated.

Most importantly, the unique role of TBP in Pol I initiation complexes remains mysterious. In a sequestered (inactive) position, TBP does not contact promoter DNA when bound by UAF31, although the factors cooperate functionally29. Furthermore, TBP does not participate in basal transcription initiation31,78 and must therefore act earlier in the initiation process. Is a role in bridging between UAF and CF sufficient to explain the effects of TBP addition? Or is there still more to learn about this factor common to all transcription systems?

Comparable to the adaptation of Pol I to synthesis of pre-rRNA, the Pol III machinery structurally and functionally adapted to the transcription of short, structured RNAs (reviewed in55,94,95). Cryo-EM structures of Pol III initiation complexes uncovered many similarities with Pol II86,87, whereas Pol I initiation mechanisms strongly differ77,96. Nevertheless, Pol III activity in yeast is mainly regulated through the repressor protein Maf1 (refs. 97,98), which responds to mammalian target of rapamycin (mTOR) signalling99,100, similar to Rrn3 in the Pol I system.

Regulation of Pol I relies on post-translational modifications in response to external cues

The mTOR signalling pathway provides a regulatory hub linking a multitude of signals with gene expression programmes to regulate and coordinate complex processes such as growth, nutrient responses or ageing (reviewed in101). Among the effects of mTOR signalling is the upregulation of Pol I transcription through mTOR complex 1 (mTORC1)-dependent activation of the kinase S6K1 in response to increased nutrient availability. S6K1 activation results in phosphorylation of the initiation factors RRN3 (ref. 102) and UBF103 in human cells and Rrn3 in yeast cells104. Simultaneously, the increased demand for nucleotides owing to enhanced Pol I activity is satisfied by mTORC1-dependent stimulation of de novo purine105,106 and pyrimidine107 synthesis pathways, and energy sources are supplied by activation of genes involved in glycolysis and the pentose phosphate pathway108. Similarly, Pol I activity can be enhanced in response to growth stimuli109, which are transduced by the MAPK signalling pathway and also lead to Rrn3 phosphorylation. In response to growth factor signalling in mouse cells, Ras–GTP activates MAPK signalling, resulting in nuclear import of ERK, which in turn allows the (direct or indirect) phosphorylation, and thus activation, of Rrn3 (ref. 110).

These examples demonstrate that Pol I regulation is most efficiently achieved by phosphorylation-dependent regulation of RRN3 association. This Pol I-specific initiation factor is defined across species by a conserved HEAT-repeat fold at its core111, which binds to the Pol I ‘stalk’ and ‘dock’ subdomains (Supplementary Fig. 1) to prevent Pol I inactivation and to stabilize a partially expanded conformation in yeast67,112 and humans113. This conformation is required for productive initiation, likely by supporting promoter DNA melting. Whereas phosphorylation of some Rrn3 residues promotes its association with Pol I, others may be inhibitory102,114. One cluster of serine residues was found to be positioned on the interface with Pol I stalk subunit A43, thus explaining that their phosphorylation abolishes Pol I interaction owing to steric hindrance and reducing rDNA transcription initiation112. A single phosphomimetic mutation of yeast Rrn3 (ref. 111) or a small peptide mimicking the central interaction site of human Pol I with Rrn3 is sufficient to shut down transcription initiation115, demonstrating why modulation of this interaction interface in cells results in efficient Pol I transcription control.

Pol I itself can be regulated by phosphorylation of two of its subunits116. In yeast, phosphorylation of the stalk subunit A43 (by an unknown kinase) supports Rrn3-dependent formation of the initiation complex117, and dephosphorylation of Pol I by the phosphatase CDC14 contributes to the shutdown of rDNA transcription during mitosis118. However, mutation of phosphorylation sites on Pol I showed only minor effects on yeast viability, indicating that their phosphorylation has a less important role in transcription regulation than Rrn3 phosphorylation116.

Taken together, recent structural and functional analyses in vivo and in vitro as well as evolutionary analyses (Box 1) indicate that target gene-specific adaptations to nucleolar Pol I transcription are advantageous, whereas general regulation of Pol II and Pol III is most effective for the transcription of all other genes.

Epigenetic regulation of transcription

The presence of non-transcribed repeat units is a major characteristic of rDNA in eukaryotes, which is counterintuitive considering a huge demand for ribosomes. A hypothesis explaining this is that silencing offers a mechanism to keep the number of active rDNA copies despite the varying rDNA copy number in the cell42,43,119. DNA CpG methylation is a major epigenetic mechanism of repression of mammalian rRNA gene transcription120,121. The effect of CpG methylation is well studied in Pol II-mediated transcription, and it is generally believed that promoter DNA methylation is associated with heterochromatin formation and transcription repression, whereas methylation at gene bodies is observed in genes with a high level of transcription122,123. In Pol I-mediated transcription, CpG methylation in the promoter region is hypothesized to inhibit transcription, and the methylation of certain cytosine positions in the rRNA gene promoter are known to inhibit binding of the transcription factor UBF121,124. However, a full picture of rDNA methylation has long been unclear as attempts at exhaustive assessment are hampered by the repeat structure of rDNA clusters. The development of long-read sequencers and super-resolution microscopy has recently enabled researchers to obtain a comprehensive view of rDNA conformation and methylation in human and mouse cells42,119,125,126.

In human cells, the rDNA coding region is methylated in an ‘all-or-none’ manner, meaning that a subset of rRNA genes is methylation-free from the promoter to the gene body, whereas other subsets are highly methylated, likely indicating that the latter are transcriptionally inactive42,125 (Fig. 4a). It was also shown that neighbouring rDNA copies are likely to have the same transcriptional state; thus, active and inactive rDNA copies form separate clusters and inactive clusters are likely to form heterochromatin42,119. The average size of such clusters was estimated to be ~20 copies, which is smaller than the average copy number of rDNA per chromosome, indicating that transcription status is not determined at the chromosome level42,127. Interestingly, the proportion of methylation-free rRNA genes is inversely correlated with rDNA copy number both in humans and in mice, suggesting a mechanism to precisely regulate the number of active rRNA genes42,128. In mice, nanopore sequencing analysis indicated that sequence variations in the promoter region are strongly correlated with the methylation frequency of the corresponding rDNA region, suggesting that rDNA transcription is determined at the sequence level119. However, a similar association between promoter sequence variations and methylation status was not found in humans, suggesting different regulatory mechanisms. Furthermore, mouse rDNA is found on the long arms of chromosomes, has a lower rate of interchromosomal rDNA recombination and has a lower level of IGS methylation compared with human rDNA clusters42,129.

Fig. 4: Regulation of rDNA transcription by heterochromatin formation through DNA methylation and histone modifications.
figure 4

a, Schematic organization of ribosomal DNA (rDNA) transcription and methylation state in the nucleolus. Inactive rDNA copies are characterized by CpG methylation in the coding region and form heterochromatin. Actively transcribed rDNA copies have reduced methylation levels and are thought to assemble in clusters of 20 tandem rDNA copies on average. The dense fibrillar component (DFC) and the fibrillar centre (FC) are regions of the nucleolus in mammals, which together form a module (see Fig. 5). RNA polymerase I (Pol I)-mediated transcription occurs in the DFC–FC interface, which probably contains one or two transcribed ribosomal RNA (rRNA) gene repeats126. b,c, Regulation of mouse rDNA transcription through DNA and histone modifications and nucleosome remodelling. b, At transcribed rDNA copies, nucleosomes are thought to be depleted between the enhancer repeats and the terminator sequence, allowing a high level of RNA synthesis. By contrast, at constitutively inactive rDNA copies, heterochromatin is formed throughout the gene cluster. The transcription factor upstream binding factor (UBF) binds to the rRNA gene body and the enhancer repeat. Transcription termination factor 1 (TTF1) is responsible for transcription termination, but also has a key role in silencing of rRNA genes through binding to T0 and Tsp sites located near the enhancer repeats, and recruiting chromatin modifying and remodelling complexes. c, Chromatin modifying and remodelling complexes act on the rDNA promoter to silence rRNA genes  and include the nucleolar remodelling complex (NoRC; constitutive silencing), the nucleosome remodelling and deacetylase (NuRD) complex and the energy-dependent nucleolar silencing complex (eNoSC) (both transient silencing). The complexes engage in adding repressive histone modifications such as histone H3 Lys27 (H3K27) trimethylation or H3K9 dimethylation, removing active-gene modifications such as histone H4 acetylation, and repositioning nucleosomes. Constitutive silencing by NoRC involves DNA methylation by DNA methyltransferases (DNMTs) and stable heterochromatin formation. Transient silencing is often the consequence of nutrient starvation and is signalled through mammalian target of rapamycin (mTOR). CSB, Cockayne syndrome B protein; HDAC, histone deacetylase; IGS, intergenic spacer; MBD, methyl-CpG-binding domain protein; pRNA, promoter RNA; TIP5, TTF1-interacting protein 5.

Full size image

A major characteristic of Pol I transcription is that nucleosomes are depleted from rDNA loci between enhancer repeats and termination sites of active rDNA copies, whereas Pol II transcribes in the presence of histones130,131 (Fig. 4b). In mouse rDNA, enhancer repeats are located upstream of the promoter and consist of an ~140-bp unit. They are flanked by an rDNA promoter and spacer promoter and by two TTF1-binding motifs termed ‘Tsp’ and ‘T0’, which are crucial for transcription regulation. Pol II transcription induces only transient dislocation of histone H2A–histone H2B dimers, and histones are retained in the gene body132. This difference could be due to the extremely high level of occupancy of rDNA by the transcription machinery. Nonetheless, histone modifications are important in rDNA transcription regulation. Histone H3 Lys9 (H3K9) methylation, H3K27 trimethylation and histone H4 Lys16 (H4K16) and H4K20 deacetylation in the promoter region are hallmarks of silent rDNA copies131,133,134,135. In active rDNA copies, H3K4 methylation, H3 and H4 acetylation and the histone variant H2A.Z are strongly enriched in nucleosomes upstream of the enhancer region131. Overall, similar histone and DNA modifications responsible for Pol II regulation are also involved in Pol I regulation in the enhancer region.

To constitutively inactivate a rDNA repeat unit, heterochromatin is established at rDNA by the nucleolar remodelling complex (NoRC), which is a chromatin remodelling complex containing the ATPase SNF2, TTF1-interacting protein 5 (TIP5) and a so-called promoter RNA136 (Fig. 4c), the last of which is a non-coding RNA produced from the spacer promoter located around 2 kb upstream of the rRNA promoter, which is required to guide TIP5 to rRNA gene promoters. TIP5 also interacts with TTF1 bound to the Tsp and T0 terminator motifs flanking the enhancer repeats (Fig. 4b).

In addition to the constitutive silencing by NoRC, rDNA transcription can be transiently repressed by different chromatin remodelling complexes in a nutrient-dependent manner2. Under glucose starvation, the histone deacetylase SIRT1, a yeast Sir2 homologue, is recruited to rDNA by the rRNA-binding protein NML to form a complex known as the ‘energy-dependent nucleolar silencing complex’ (eNoSC)137,138 (Fig. 4c). eNoSC deacetylates H3K9 at the rDNA promoter and induces its dimethylation, thereby establishing transient rDNA silencing. In addition to its role in RRN3 phosphorylation during Pol I inhibition, the mTOR pathway also regulates eNoSC activity139,140,141. Another important factor is the nucleosome remodelling and deacetylase (NuRD) complex, which is composed of histone deacetylases, methyl-CpG-binding domain protein 3 (MBD3) or MBD2, the ATP-dependent helicase CHD4, and MTA 1, MTA2 or MTA3, and is suggested to work as a scaffold142. The NuRD complex is recruited to the rDNA enhancer by interacting with TTF1 and Cockayne syndrome B protein (CSB) and establishes a transient and reversible repressive rDNA state.

Unlike the constitutive transcription regulation by NoRC, the nutrient-dependent repression mechanisms do not appear to cause DNA methylation143. Another difference in mammals between temporary repression of rDNA repeats and their constitutive silencing is the chromatin accessibility of the gene body. Whereas constitutively silenced rDNA copies are not susceptible to intercalation by the DNA crosslinking agent psoralen, indicating chromatin compaction, temporarily silenced copies are sufficiently accessible to allow psoralen intercalation, which indicates that temporary repression is mainly achieved through mechanisms that limit transcription initiation rather than inducing full heterochromatin formation. In summary, rDNA silencing is achieved through DNA methylation, histone modification and chromatin remodelling complexes in reaction to environmental cues to rapidly adapt rDNA transcription levels. Nevertheless, a detailed understanding of the molecular mechanisms that distinguish between constitutive and transient rDNA repeat silencing and reactivation remains to be established.

Nucleolus architecture and function

The nucleolus is a membraneless compartment of the nucleus. It is the centre of rRNA synthesis and processing, and differs in its architecture among yeast, amphibians and mammals. In the following sections, we discuss the architecture of the nucleolus and the role of phase separation in rDNA transcription and its regulation, as well as challenges in the field.

Architecture of the nucleolus

The nucleolus of vertebrates is a three-layer structure composed of the granular component (GC), the dense fibrillar component (DFC) and the fibrillar centre (FC)144. Transcription occurs in the interface between the FC and the DFC, indicating that active rRNA genes may form loops126 (Fig. 5a). Such loop formation is proposed to increase the rate of reloading of Pol I onto the rDNA promoter by the juxtaposition of the promoter and the terminator elements145. Once transcribed, pre-rRNAs are released into the DFC region, where the bulk of pre-rRNA processing occurs125,146 (Fig. 5b). Finally, within the GC, assembly of mature rRNAs with ribosomal proteins and incorporation of 5S rRNA (transcribed by Pol III from distinct loci outside the nucleolus) occurs147. The so-called Miller spread technique was developed to visualize transcribed rDNA and revealed a ‘Christmas tree’-like structure, in which the ‘trunk’ corresponds to Pol I-decorated rDNA, the ‘branches’ are the nascent rRNAs and ‘Christmas balls’ are early ribosome assembly intermediates56 (Fig. 5b). Detailed analysis of such ‘Miller trees’ helped our understanding of different aspects of rDNA transcription93,148,149, including the effect of rDNA copy number mutations on Pol I occupancy of rDNA, and showed that ‘Christmas balls’ contain the rRNA processing complex, indicating that rRNA processing is already occurring co-transcriptionally150.

Fig. 5: Functional organization of the nucleolus.
figure 5

a, Schematic representation of a mammalian nucleolus in an unperturbed nucleus (left) and a nucleus with inhibited RNA polymerase I (Pol I) activity (right). The nucleolus is composed of three layers — the granular component (GC), the dense fibrillar component (DFC) and the fibrillar centre (FC) — and ribosomal DNA (rDNA) transcription occurs at the interface between the FC and the DFC. In super-resolution microscopy images, the DFC appears to consist of multiple ‘beads’ that surround an FC core125. When transcription is inhibited, mammalian nucleoli undergo a characteristic change called ‘nucleolar segregation’. In such a segregated state, nucleoli become spherical, and the transcription machinery accumulates at the periphery of the GC, forming structures called ‘nucleolar caps’. b, Nucleolar architecture and the organization of ribosomal RNA (rRNA) synthesis. Transcription of rRNA genes occurs between the FC and the DFC, and the nascent pre-rRNA is sorted into the DFC125. Visualized as chromatin spread (Miller spread) by electron microscopy, the transcribed rDNA and pre-rRNAs form a tree-like structure, in which progressively longer rRNAs and ribosome intermediates form as Pol I transcribes along rDNA56. Processing and modification of pre-rRNAs occur mainly in the DFC, and their assembly with ribosomal proteins (RPs) and Pol III-synthesized 5S rRNA occurs in the GC144. The nucleolus is surrounded by heterochromatin, which serves as a nuclear domain for loci that form constitutive heterochromatin such as satellite repeats. Even though some inactive nucleolus organizer regions (NORs) have been observed outside nucleoli, most NORs are associated with nucleoli in most cell types, and inactive rDNA copies appear to form heterochromatin inside the GC127.

Full size image

The nucleolus organizer regions, which contain all rDNA repeats, are mostly associated with nucleoli, although nucleolus organizer regions can also be found elsewhere in the nucleus42,127 (Fig. 5b). Association with nucleoli is suggested to be achieved through repeats at the boundaries between rDNA and non-rDNA regions127,151. Most of the inactive rDNA copies are located inside nucleoli and are thought to form heterochromatin inside the GC. Electron microscopy studies have shown that whereas heterochromatin comprises a small volume inside nucleoli, the outer periphery of nucleoli is typically also rich in heterochromatin, which is estimated to be formed by non-rDNA regions such as telomere and centromere as well as by rDNA149,152,153.

The architecture of the nucleolus is known to diverge not only between organisms but also between different cell types. This likely reflects different transcriptional activity of rDNA and ribosome requirement. For example, mouse embryonic stem cells typically contain a single large nucleolus, whereas in differentiated cells, several smaller nucleoli are observed154,155. This is thought to reflect the hyperactive nature of mouse embryonic stem cell rDNA42,154. Neurons, which are more transcriptionally active than most differentiated cells, are also known to harbour larger nucleoli, and nucleoli of postnatal neurons become increasingly larger and reticulate as the neurons mature156,157.

Liquid–liquid phase separation in the nucleolus

How can membraneless compartments such as the nucleolus be stably and clearly separated from the rest of the nucleoplasm? Following highly influential work which described liquid-like behaviour of germ line P granules in Caenorhabditis elegans158, liquid–liquid phase separation (LLPS) has emerged as a concept potentially explaining the formation of distinct protein and RNA condensates in cells144. Condensates formed through LLPS show liquid-like characteristics such as high mobility of molecular components and fusion between droplets. LLPS is thought to be driven by RNAs and by intrinsically disordered regions of proteins, which assemble through weak and multivalent interactions. LLPS is thought to support high local concentrations of phase-separated factors, thereby enabling acceleration of biochemical processes and fast adaptation to environmental changes159,160.

It was later also found that extrachromosomal nucleoli in Xenopus laevis oocytes, which are formed by extrachromosomal amplification of rDNA (Fig. 1b), have characteristics of LLPS159. X. laevis oocyte nucleoli are spherical and can fuse when experimentally forced to contact each other. ATP depletion causes these nucleoli droplets to lose their viscosity, suggesting that active ribosome production underlies the liquid-like properties of X. laevis nucleoli. Purified nucleolar proteins form droplets in vitro and mixing of different nucleolar proteins recapitulate features of X. laevis oocyte nucleoli, such as the multilayered structure composed of the GC and the DFC161. The mechanism behind this droplet formation is thought to be driven by intrinsically disordered regions of nucleolar proteins (nucleophosmin 1 for the GC and fibrillarin for the DFC) and by nascent rRNA161,162. Pol II transcription is also suggested to be organized in condensates driven by LLPS (reviewed in163). Formation of these condensates relies on the C-terminal domain repeats of the largest subunit of Pol II, which can be heavily modified and is thought to serve as an intrinsically disordered region164,165,166. However, Pol I does not have such a domain, suggesting that if Pol I undergoes or promotes LLPS, it does so through a mechanism different from that of Pol II.

Amniotes, including mammals, reptiles and birds, have a nucleolus architecture different from that of anamniotes, invertebrates and plants167. Whereas amphibian nucleoli harbour multiple FCs within each DFC, the DFC and FC of mammalian nucleoli form a two-layered module, and several dozen of these modules are included in the GC (Fig. 5a). This architecture may result from an evolutionary increase in the length of the IGS segments in mammals, as the establishment of the DFC–FC modules coincides with the acquisition of longer IGS segments167. The shape of mammalian nucleoli is also not typically spherical as would be expected from a model of simple LLPS, but is irregular. In addition, nucleolar size and shape differ considerably between cell types126,168. Super-resolution microscopy studies have revealed that the mammalian DFC consists of multiple bead-like protein aggregates, which are located so close to each other that they should theoretically fuse with each other if they behaved liquid-like125 (Fig. 5a). Finally, inhibition of rRNA transcription was found to result in a characteristic structural change termed ‘nucleolar segregation’. This segregation is characterized by a fusion of the DFC and the FC, which contains the rDNA and its transcription machinery. The fused layers and the transcription machinery then migrate to the periphery of the nucleolus, forming nucleolar caps, and the GC is transformed to a spherical shape169 (Fig. 5a). In the nucleolar caps, Pol I proteins in the FC phase are more mobile and likely not bound to rDNA compared with their mostly immobile state in normal nucleoli170. Therefore, liquid droplet-like behaviour of nucleoli is increased by segregation of nucleoli triggered by transcription repression. The opposite is the case in amphibian oocytes, which require active transcription to maintain liquid-like properties. Moreover, mammalian nucleoli do not have a fixed saturation concentration, in contrast to single-component LLPS artificially realized using purified proteins, meaning that their thermodynamic status are also different162,171,172. For these reasons, the mammalian nucleolus does not strongly resemble LLPS-driven protein–RNA condensates in vivo or artificial in vitro droplets made from a single component.

In conclusion, although the exact physical mechanisms and functional contribution of LLPS to the formation of nucleoli are still unknown, these findings have advanced our understanding of the nature of nucleoli. Future research should also address why nucleolar organization diverges between species and cell types and what functional impact these differences have.

rDNA, cell senescence and disease

As the rRNA gene is the most abundant gene and the most highly transcribed, its stability and transcription rate affect cellular functions on many levels. For example, the rDNA locus in budding yeast shows such high recombination frequencies that its instability may influence cell senescence5. Senescence may be triggered when rDNA recombination produces ERCs (Fig. 1b), which contain a replication origin and thus replicate in every cell cycle. As ERCs asymmetrically segregate to the yeast mother cell35, their abundance exponentially increases in the mother cell and they sequester factors important for chromosome maintenance, such as histones, resulting in stagnating cell growth and finally death173,174,175. ERC accumulation is not observed in mammalian cells; therefore, the described ERC-dependent senescence seems to be a budding yeast-specific phenomenon. A recent study of a yeast strain producing fewer ERCs owing to partial reduction of replication initiation within the rDNA cluster revealed increased rDNA instability due to replication stress and shorter cellular lifespan, indicating that rDNA instability itself may promote senescence independently of ERC accumulation176,177,178. Strikingly, considerable rDNA instability is detected in old mother cells, but not in the daughter cells, suggesting that, in budding yeast, the rDNA produces an ‘ageing signal’, which reflects the level of genome instability in a cell176,178,179.

A correlation between rDNA stability and cell senescence was also suggested in mammalian cells. Long-read DNA sequencing-based analysis revealed that the proportion of irregular rDNA units, such as those with large deletions, inversions and so on, is increased eightfold in cells from persons with progeroid syndromes such as Werner syndrome and Bloom syndrome42, where mutations in DNA repair genes lead to accelerated ageing and shorter lifespans. As rDNA is unstable in nature (see earlier) and about half of the units are not transcribed at any given time, DNA damage tends to accumulate at these sites and triggers a cell senescence mechanism180. Pronounced rDNA instability characterizes cancer cells with increased genomic instability, which could have arisen from cells escaping senescence181,182.

In addition to potentially contributing to ageing, accumulation of rDNA damage itself reduces the quality of rRNA and thus of ribosomes, which may also induce an ageing phenotype. Mutations in the 28S rRNA gene were found specifically in old mice, and their introduction to yeast rDNA shortened yeast lifespan183. In addition to mutations, epigenetic regulation of rRNA genes also changes with age, and old mouse and human cells have higher rDNA methylation levels than young cells, and the number of their pre-rRNA transcripts is reduced183,184,185. Such an age-dependent decrease of the quality and quantity of pre-rRNAs is likely to reduce cell function and accelerate cell senescence.

Pol I activity also determines rRNA quantity, and human Pol I subunit mutations are linked to at least four developmental diseases: acrofacial dysostosis (Cincinnati type)170,186, hypomyelinating leukodystrophy (HL), Treacher Collins syndrome187,188,189,190 and a juvenile neurodegenerative phenotype akin to the HL phenotype that is characterized by severe neurological deficits in childhood or adolescent development191. Mapping of these mutations to human Pol I structures allowed speculation about the molecular mechanisms underlying the pathologies76,113,192. Essentially, the mutations described to date are likely to cause problems in Pol I translocation during nucleotide addition, minor subunit folding defects, or intersubunit assembly or stability defects. Although acrofacial dysostosis (Cincinnati type) and the juvenile neurodegenerative phenotype similar to HL are caused by mutations in Pol I-specific subunits, HL-related and Treacher Collins syndrome-related mutations can occur in the subunits shared by human Pol I and Pol III. Interestingly, causative mutations in HL cluster in the C-terminal region of the subunit RPAC1, which is a structured region in human Pol III193,194,195,196, but not in human Pol I, indicating that the main effects may originate from Pol III defects. Similarly, rDNA stability and transcription are heavily related to the cellular function.

Conclusions and future perspective

The primitive organisms that existed before cells as we know them today appeared are speculated to have looked like a ‘droplet’, somewhat comparable to the current nucleolus without rDNA197. As protein-coding genes require ribosomes for their expression, rRNA and tRNA genes may be the first genes to have evolved. In this Review, we have discussed how rDNA maintenance and transcription have evolved to support the unique requirements in quality and quantity of rRNA production. Although we focused on E. coli, budding yeast and human (and mouse) rDNA, other organisms, such as plants, have conserved similar systems198.

The status of rDNA is deeply intertwined with cell senescence and it restricts cellular lifespan. This vulnerability of rDNA instability may have contributed to evolution by limiting alteration of generation times. In other words, by restricting cellular lifespan, rDNA instability seems to drive evolution. It will be interesting to analyse the relationship between rDNA stability and lifespan in many organisms. Alterations of rDNA stability by improving repair capacity or inhibition of recombination may indeed influence the lifespan of complex organisms as observed in budding yeast. If so, manipulating rDNA stability could be a promising strategy to increase life expectancy, although it might simultaneously enhance issues that more frequently appear with ageing, such as cancer and dementia. Overexpression of a mammalian Sir2 homologue (SIRT6) is known to extend lifespan, and the activation of homologues by the coenzyme NAD+ (or its precursors) promotes fitness in mice199,200. Although there is no evidence that rDNA is actually stabilized in these mice, a similar effect may be expected in humans201. In addition, it seems that Pol I hyperactivity is observed in many cancers, and inhibition of Pol I transcription may even be used as a therapeutic strategy (reviewed in202).

In conclusion, through the maintenance and regulation of ribosome production, rDNA has the strongest impact on the life and evolution of all organisms, as it comprises the most abundant, most unstable, most important and oldest gene.